The model provides prediction probability to the user before submitting an emulation job to high volume product manufacturing. The user can decide to modify ingredients in a job with low likelihood of passing the tests. The test results are automatically loaded to the database, and the model is retrained using updated results in the database. This approach save emulators’ time spent on running jobs, which are highly unlikely to pass the tests, without incurring extra efforts to the users. In this paper, we will discuss the implementations using Random Forest and Neural Network, evaluate their performances, and discuss considerations for continuous deployment.